Psychological Medicine, 2003, 33, 13691379.

f 2003 Cambridge University Press

DOI : 10.1017/S0033291703008390 Printed in the United Kingdom

Examining the relationship between pubertal stage,

adolescent health behaviours and stress
A. E. S I M O N, J. W A R D L E,1 M. J. J A R V I S, N. S T E G G L E S AND M. C A R T W R I G H T
From the Cancer Research UK Health Behaviour Unit, Department of Epidemiology and Public Health,
University College London

ABSTRACT
Background. This paper examines the associations between puberty and three important health
behaviours (smoking, food intake and exercise) and explores whether these associations are
mediated by pubertys relationship to stress and psychological diculties.
Method. Data were taken from the rst year of the ongoing, 5-year, Health and Behaviours in
Teenagers Study (HABITS). This is a school-based study set in 36 schools in London. In the rst
year of the study, 4320 students (2578 boys, 1742 girls) in their rst year of secondary education
took part.
Results. Among girls, being more pubertally advanced was associated with a greater likelihood of
having tried smoking. Among boys, being more pubertally advanced was associated with a greater
likelihood of having tried smoking, a higher intake of high-fat food and higher levels of exercise.
More pubertally advanced girls experienced more stress but not more psychological diculties.
There were no associations between puberty and either stress or psychological diculties in boys.
Stress and psychological diculties were associated with health behaviours in girls and boys, but
neither of these factors mediated the relationship between pubertal stage and health behaviours
found in girls.
Conclusions. These results suggest that the onset of puberty has a marked eect on the development
of health behaviours. Puberty was related to an acceleration of the development of unhealthy
behaviours, except for exercise behaviour in boys, where advanced puberty was associated with
more exercise. These changes were unrelated to adolescent issues of stress and a causal explanation
for these associations must be sought elsewhere.

INTRODUCTION including varied sleeping and eating patterns

(Gillman et al. 2000 ; Laberge et al. 2001), new
Puberty is a period of rapid hormonal, physio-
sexual behaviours (Goodson et al. 1997) and
logical and physical changes. These dramatic
involvement in delinquency and substance use
changes might be expected to engender some
(Williams & Dunlop, 1999; Dick et al. 2000).
psychological eects, insofar as they represent a
The hormonal and physical changes of pu-
developmental transition for which the individ-
berty will have a direct inuence on some aspects
ual is not always fully prepared. Onset of pu-
of emotion and behaviour. In addition, the per-
berty has been associated with a range of social
ception of changes by the adolescents themselves,
and emotional eects (Steinberg & Sheeld-
Morris, 2001). Behaviour changes also appear, or the reactions of others to their more adult
bodies, could play a part. There is evidence
that the timing of puberty inuences the individ-
1
Address for correspondence: Professor Jane Wardle, Cancer uals reactions. Being pubertally o-time , i.e.
Research UK Health Behaviour Unit, Department of Epidemiology
and Public Health, University College London, 216 Torrington early or late compared to others in the social
Place, London WC1E 6BT. group, has been hypothesized to be stressful
1369
1370 A. E. Simon and others
(Petersen & Taylor, 1980). This is partly because
the o-time adolescent is in the minority and
may lack the social or institutional support that
smooths the transitional path for the on-time
individuals (Alsaker, 1996). The stress of o-time
puberty has been hypothesized to lead to adapt-
ive problems, which display themselves in the
form of emotional problems or delinquency
(Williams & Dunlop, 1999). An alternative
hypothesis is that only early maturers are at
particular risk because their physical maturity
means that both adults and peers attribute to
them a psychological maturity that may outstrip
their actual development (Stattin & Magnusson,
1990). Early maturing girls tend to develop
friendship networks with other equally mature,
and therefore older girls, where more adult-
type behaviour is acceptable (Silbereisen et al.
1989). Early maturers may also perceive them-
selves as more grown-up and emulate adult
behaviour, such as smoking and dieting. A con-
ict can arise because these behaviours, which
are acceptable within the adult world, are per-
ceived by adults to be norm-breaking or rule
transgressing in teenagers (Alsaker, 1996).
Behavioural research in this eld has had a
particular focus on behaviours such as sub-
stance abuse and delinquency (Tschann et al.
1994 ; Williams & Dunlop, 1999 ; Dick et al.
2000), which pose problems in the adolescent
years, compromising family and social relation-
ships, as well as educational progress. Delin-
quent behaviour peaks during mid-adolescence,
then declines to pre-adolescent levels by the
mid-twenties. This increase in delinquency may
reect a maturity gap between biological
maturity and access to adult roles in society
(Mott, 1996). As regards substance use, studies
have shown that early-maturing adolescents,
but not all o-time adolescents, have higher
substance use, including alcohol, marijuana and
cigarette use (Tschann et al. 1994; Harrell et al.
1998). Early menarche has also been associated
with earlier and more frequent smoking and
drinking (Dick et al. 2000).
Smoking has most often been examined in
this broader context of substance use or in the
delinquency context where it is represented as a
transgressing act . However, adolescence is also
an important time for establishing health behav-
iours, and it is possible that the onset of puberty
could be linked with a higher risk not only of
smoking, but also of other health-compromising
behaviours such as a sedentary lifestyle and an
unhealthy diet. Early initiation of unhealthy
behaviours might place the individual at higher
risk for health problems in later life.
Research on exercise in adolescence shows
that girls participation in exercise is consistently
lower than boys participation (Lindquist et al.
1999 ; Lasheras et al. 2001). Boys interest in
school-based physical education (PE) remains
relatively stable throughout secondary school,
but girls interest in PE, while higher than boys
at younger ages, drops o, so that by the age of
14 the gender balance is reversed (Van Wersch
et al. 1992). It has been suggested that PE loses
its status as an important school subject for
girls, because being good at sports does not
entail popularity for girls in the way that it does
for boys (Van Wersch et al. 1992). Boys have
also been shown to use exercise rather than
dieting as their primary route for body shape
change (Ricciaridelli et al. 1999). On the basis of
these results, being more pubertally advanced
might be expected to be associated with higher
levels of exercise in boys and lower levels in girls.
One study, which examined pubertal stage in
relation to activity choices, found that higher
pubertal stage was associated with an increase in
sedentary activities in girls, although there was
no such association in boys (Bradley et al. 2000).
Energy intake has been investigated in re-
lation to pubertal stage in a number of studies,
but there are few studies of food choices. Total
energy intake increases across pubertal stage,
peaking for girls at Tanner stage 3, and con-
tinually increasing with stage for boys (Clavien
et al. 1996), although after controlling for body
weight, the more pubertally advanced ado-
lescents are not consuming more in relation to
their body weight. Adolescence is a time when
children take more responsibility for their own
food choices, and are more likely to eat at dif-
ferent times from the family or away from the
family home. Adolescents also appear to con-
sume an excess of fat at the expense of carbo-
hydrate and bre intake (Clavien et al. 1996;
Rolland-Cachera et al. 2000), and eat little fruit
and vegetables (Neumark-Sztainer et al. 1996)
although this has not been examined extensively
in terms of pubertal development.
Some of the work relating pubertal stage
to behavioural outcomes has considered the
 Puberty and adolescent health behaviours
possibility that emotional factors provide the
link with pubertal stage (Tschann et al. 1994;
McCabe et al. 2001). Puberty has been shown to
be related to depression, antisocial behaviour,
aggression, anxiety, withdrawal and attention
problems (Caspi & Mott, 1991; Hayward et al.
1999). In addition, the increase in weight at pu-
berty leads to body dissatisfaction and body
change behaviour, such as dieting, that are also
related to increased levels of anxiety (McCabe
et al. 2001). However, few studies have directly
tested the mediational model. One exception is a
study by Tschann et al. (1994), which found that
emotional distress did not appear to mediate the
relationship between pubertal timing and sub-
stance use. This studys focus was an examin-
ation of substance use and it would be useful to
both broaden the exploration of this topic to
other kinds of health behaviours and to provide
further evidence either to replicate or refute this
nding.
The present study examines associations be-
tween puberty and three signicant health be-
haviours : smoking, food choice and exercise
and examines the hypothesis that emotional
stress mediates the associations between puberty
and behaviour. In this way, a fuller picture
can be developed of the relationships between
pubertal development, health behaviour and
emotional distress. Data are taken from the
baseline year of a 5-year cohort study, giving the
advantage of a large, mixed-sex sample (Wardle
et al. 2003).
METHOD
Design
The HABITS study is a 5-year longitudinal
study of a cohort of adolescents attending 36
secondary schools in Greater London. The
sample was stratied by school type (inner-city
state ; suburban state ; privately funded) and
gender mix (boys, girls, both). A list was com-
piled of all secondary schools in the South
London boroughs taking students from the age
of 1116 years. Four schools were drawn at
random from each of the nine cells formed by
crossing the stratication factors. Each school
was approached. If a school declined to take
part a substitute was drawn from the list until
all cells were complete. Twenty-eight schools
1371
declined to take part. Thirty-six schools who
agreed to participate were identied. The data
for this paper were from the rst year of
data collection between January and December
1999.
All students in the rst year (age 1112 years)
of the secondary schools were eligible for in-
clusion (N=5153). Parental approval was ob-
tained prior to data collection in the form of
a letter taken home from school by students.
Parents returned forms to the school if they
wished to exclude their child from the study. In
the rst year 4320 students (2578 boys and 1742
girls) in Year 7 took part. Eighty-four per cent
of the total student body were present in the
classes when data collection took place. Ten per
cent were absent on the day of data collection,
and 5.5 % opted out of the study (either by
parental or self-exclusion).
Participants completed a self-report question-
naire on smoking, diet, exercise, body image,
pubertal development, stress, psychological
health, and personality. Demographic infor-
mation included ethnicity, gender, and socio-
economic status (SES). Smoking status was
conrmed with a saliva cotinine test. Body size
was assessed with height, weight, and waist
circumference. Ethical approval for the study
was obtained from the Joint UCL/UCH Com-
mittees on the Ethics of Human Research.
Procedure
Data collection was carried out in school classes
under the supervision of at least two re-
searchers. Data were collected annually during
the Spring and Autumn terms, with 18 schools
visited each term.
A researcher introduced the session by ex-
plaining the purpose, the procedure, the con-
dentiality of answers, and the voluntary nature
of the study. An opportunity for the students to
ask questions was provided. Students completed
a sheet (with a unique identier) detailing their
name, address, school name, date and conrm-
ing their consent to participate. This was stored
separately for anonymity and used to track
students across the 5 years. Students completed
the questionnaire and were individually weighed
out of the sight of the other students. Saliva
samples for cotinine testing were obtained using
a dental roll placed in the side of the mouth for
10 min (Benowitz, 1996).
1372 A. E. Simon and others
Measures
Pubertal stage
Pubertal stage was assessed with items from the
Pubertal Development Scale (PDS) (Petersen
et al. 1988). The PDS was developed for use in
schools, where other measures such as direct
observation or explicit photographs (e.g. Sexual
Maturation Scale (SMS) ; Tanner, 1962) are in-
appropriate. The PDS scale is correlated with
the SMS (r=0.72 to 0.80) (Brooks-Gunn et al.
1987). Students rate themselves on ve items :
growth spurt, pubic hair, skin changes, and, for
girls, menarche and breast development and, for
boys, voice change and facial hair. Scores from
each item were summed (range 5 to 19). Petersen
et al. (1988) report that this measure can be
used either continuously or with a categorical,
staging classication. In this sample, a 5-stage
classication system has been used : pre-pubertal
(score 5), beginning pubertal (scores 610),
mid-pubertal (scores 1115), advanced pubertal
(scores 1618), and post-pubertal (score 19).
Health behaviours
Smoking status was assessed using two ques-
tions from the UK National Smoking Surveys
(Goddard & Higgins, 1999). The classication
from these questions is : non-smokers (never
smoked), tried smokers (tried only once), ex-
smokers (used to smoke), occasional smokers
(<1 a week), sometimes smokers (16 a week)
and regular smokers (o6 a week). For some
analyses, because cell sizes of regular smokers
at this age were small, students were categor-
ized into never smokers and ever smokers .
Susceptibility to future smoking among non-
smokers was assessed using a scale developed by
Pierce et al. (1996) that taps the pre-experimen-
tation stage of the smoking onset process. The
scale in this study is based on two items, refer-
ring to the possibility of smoking in the next
12 months and smoking a cigarette oered by
a best friend. The response options are : yes
denitely , yes probably , probably not and
denitely not. If the respondent answers in any
other way than denitely not they are con-
sidered cognitively susceptible to experimen-
tation with smoking.
Weekend exercise rates were measured by two
questions developed for the study which asked
if participants usually took part in sports or
physical activities on Saturday and Sunday. The
response options were dichotomous (yes/no).
This indicated voluntary participation in extra-
curricular sports and provides a measure of ex-
ercise independent of compulsory school sports
frequency, which shows considerable variation
between schools. This variable is likely to reect
a social aspect of exercise behaviour.
Days per week exercised was assessed using
a question developed in the Youth Risk Be-
haviour Surveillance System Questionnaire
(YRBSSQ) (Heath et al. 1993). This asks how
many days a person did exercise during the
past week that made them sweat and breathe
heavily. The response ranged from no days to
every day. This indicated overall exercise, in-
cluding participation in sports at school.
Fruit and vegetable intake was assessed using
two questions, originally based on items from
the DINE (Roe et al. 1994) and used in a simi-
lar, modied format in other studies (e.g.
Wardle et al. 2000). The questions asked : about
how many servings of fruit/vegetables do you
usually eat in a day? . Response options ranged
from less than one to ve or more servings per
day. The responses were summed to give a scale
of 010, 0 represents intake of less than one
serving per day and 10 represents consuming
ve or more servings of fruit and ve or more
servings of vegetables per day.
Intake of high-fat foods was based on six
items using a scale adapted from the foods that
you eat scale (Sweeting et al. 1994). These items
related to intake of crisps, cakes, biscuits, chips,
fried food and sausages/burgers. These items
provide proportionally high levels of fat and are
often used as snack foods outside mealtimes.
Intake was assessed on a six-point scale ranging
from never/less than once a week to more
than once a day . Scores were summed, pro-
ducing a range from 030, higher scores indicate
higher intake. This scale had good reliability
(alpha=0.78).
Stress and psychological diculties
Psychological diculties were measured using
the Diculties score from the Strengths and
Diculties Questionnaire (SDQ) (Goodman
et al. 1998). This is a brief behavioural screening
questionnaire for use with children from age
416 years. The SDQ measures hyperactivity,
emotional symptoms, conduct problems, peer
 Puberty and adolescent health behaviours
problems and also has a pro-social scale. The
pro-social scale is excluded when examining
psychological problems. Diculties scores
range from 0 to 34 with a higher score indicating
more diculties. In this sample it had good in-
ternal reliability (alpha=0.72). The SDQ has
good concurrent validity and is highly corre-
lated with scores on the Rutter questionnaires
(Goodman et al. 1998).
Stress was assessed with the 4-item version of
the Perceived Stress Scale (PSS) (Cohen et al.
1983) a global measure of perceived stress. The
scale ranges from 0 to 16, higher scores indicate
more stress. The PSS has good concurrent and
predictive validity and is correlated in the ex-
pected direction with a range of questionnaires
and behaviours (Cohen et al. 1983). This
shortened version has relatively low internal
reliability (alpha=0.6) (Cohen & Williamson,
1988), which was lower (alpha=0.50) in the
present sample, possibly because of the age of
the respondents. Its advantage is that it is short
and enables repeated measures of perceived
stress in large samples.
Demographic measures
Demographic information included gender, age,
self-reported ethnicity and SES. Students were
asked their exact date of birth, and age in deci-
mal years was calculated. Ethnicity was assessed
with ve response categories : white, black,
Asian, mixed or other (specied). The categories
mixed and other contained heterogeneous
samples and were combined for the purpose of
analyses. SES was assessed using indicators of
material conditions in the family home. More
traditional methods of assessing SES (e.g. par-
ental occupation) are unreliable in lower SES
adolescents (Wardle et al. 2002). Three material
indicators were used : housing tenure, car own-
ership and possession of a dishwasher.
Body Mass Index (BMI )
BMI was derived from height and weight
measurements.
Analytical strategy
Descriptive statistics are presented for the PDS
scale and the demographic characteristics of the
sample. Chi-square statistics and one-way
analyses of variance (ANOVA) are then shown
for the relationships between pubertal stage,
1373
demographics, health behaviours and psycho-
logical variables.
The data were then analysed using logistic
and ordinal regressions showing the relationship
of pubertal stage to health behaviours for each
gender, while also controlling for SES, ethnicity,
age and BMI. The dependent variables in the
logistic regressions were: smoking status (tried
smoking or not), susceptibility to smoking (sus-
ceptible or not) and weekend exercise (12 days
or none) and in the ordinal regressions were :
days per week exercised (07 days), fruit and
vegetable intake (010 servings) and high-fat
food intake (separated into deciles). In order to
assess the possibility that stress/psychological
diculties mediate the relationship between
pubertal stage and health behaviours, further
regression analyses were carried out which
compared odds ratios from a model that in-
cluded stress/psychological diculties and a
model without these variables. The regression
analyses were carried out in STATA v.7 and
took account of school clustering and a com-
bined weighting based on a weight for design
and missing data. The design weight was based
on the total number of schools in South London
within each stratication cell and the refusal
rate of the schools within each cell. The missing
data weight took account of the proportionately
greater amount of missing data from the lower
SES and ethnic minority groups.
RESULTS
The demographic characteristics of the partici-
pants are shown in Table 1. The average age of
the students was 11.83 decimal years (S.D.=
0.35). Approximately 40 % of the participants
considered themselves to be from non-white
ethnic groups.
Table 1 also shows pubertal stage using the
PDS scores. For both boys and girls, few are
pre-pubertal, the majority are classied as be-
ginning pubertal , with more boys than girls in
this category, and more girls represented in the
more advanced categories. This is to be expected
at this age as girls mature earlier than boys. For
most of the analyses, the sample was divided
into three groups. This was developed because
the cell sizes for boys who had reached advanced
puberty and for girls who had still not begun
puberty at this age were small. For girls the
1374 A. E. Simon and others

Table 1. Distribution of the sample by SES, either been accelerated or held back at school.
ethnicity, pubertal stage, BMI and age As would be expected, the mean age increases
across pubertal stages in girls, F(2, 1301)=
Girls
(N=1742)
Boys
(N=2578)
26.20, P<0.001 ; and boys, F(2, 1720)=13.00,
P<0.001. Students who were advanced, and
Age, mean (S.D.) 11.85 (0.32) 11.81 (0.37) thus more developmentally mature, also had a
19.98 (3.84) 19.02 (3.40)
BMI, mean (S.D.)
SES quartiles, %
higher BMI : girls, F(2, 1327)=48.50, P<0.001;
Low 25.2 18.7 boys, F(2, 1775)=8.55, P<0.001.
Medium 25.0 21.4
High 27.1 31.5 Pubertal stage and health behaviours
Highest 22.7 28.4
Ethnicity, % Table 2 also shows the raw data for the re-
Black 21.6 15.3 lationships between pubertal stage and health
White 58.1 64.3 behaviours. These indicated that, in boys, a
Asian 7.8 8.9
Mixed/Other 12.6 9.8 relationship existed between pubertal stage
Pubertal stage, % and high-fat food intake, F(2, 1719)=8.22,
Pre- 2.1 8.1 P<0.001, fruit and vegetable consumption,
53.1 73.0
Beginning
Mid- 36.0 18.1
F(2, 1758)=3.64, P<0.05, having tried smok-
Advanced 8.4 0.6 ing x2(N=1757)=26.61, df=2, P<0.001 and
Post- 0.4 0.2 weekend exercise, x2(N=1723)=15.94, df=4,
P<0.01. In girls a relationship was seen be-
tween pubertal stage and having tried smoking,
original pre and beginning categories were x2(N=1313)=31.47, df=2, P<0.001.
collapsed into a single category (pre/beginning In order to explore these relationships more
puberty) and, for boys, the original mid and rigorously a series of logistic and ordinal re-
advanced/post categories were collapsed into gression analyses were carried out to assess the
a single category ( mid/advanced ). Therefore, odds of reporting each behaviour in relation to
the three groups in girls were pre/beginning pubertal stage, controlling for age, SES, eth-
puberty , mid-puberty and advanced/post nicity, and BMI and including clustering for
puberty , whereas the three groups in boys rep- schools and weighting for design and missing
resented pre-puberty , beginning puberty data, as previously discussed. The lowest pu-
and mid/advanced puberty . All analyses are bertal stage was used as the reference group. In
carried out separately for boys and girls. the rst sets of analyses, the gender-by-pubertal
Table 2 shows the associations between stage interactions were included in the model.
demographic characteristics including ethnicity, These were signicant or close to signicant, for
SES, age and BMI and pubertal stage. There weekend exercise, days of exercise, high-fat food
was a signicant relationship between SES and intake, and fruit and vegetable intake, although
pubertal stage in boys, x2(N=1715)=18.92, not for either of the smoking variables. In the
df=6, P<0.01 and a similar trend in girls, light of these results, and because of the dierent
x2(N=1316)=11.64, df=6, P=0.07, with the groupings of pubertal stage in boys and girls,
students from lower SES backgrounds being the analyses of the associations between pu-
more likely to be advanced maturers. Among bertal stage and health behaviours were carried
both boys and girls, the black students were out separately for each gender. These results are
more likely to be in the more advanced pubertal shown in Tables 3 and 4.
stages, compared with children from white and The odds of smoking were higher in the more
Asian backgrounds : boys, x2(N=1770)=22.91, pubertally advanced students in both boys and
df=6, P<0.001 ; girls, x2(N=1327)=76.54, girls. The pattern of odds ratios supported a
df=6, P<0.001. graded eect across the stages, indicating that
The age range for Year 7 is 10.3 to 13.7 deci- relative to students in the earliest pubertal stage,
mal years. The typical age of students in Year 7 those at successively later stages were increas-
is 1112 years, the sample also included nine ingly likely to have smoked. There was also a
participants aged <11 and 10 participants relationship between pubertal stage and non-
aged >12. These represent children who have smokers susceptibility to smoking in both boys
 Puberty and adolescent health behaviours 1375

Table 2. Descriptive statistics showing the relationship between pubertal stage, demographics, health
behaviours, stress and psychological diculties
Girls Boys

Pubertal stage Pubertal stage

Pre/Beginning Mid Advanced/Post Pre Beginning Mid/Advanced

(N=739) (N=482) (N=118) (N=145) (N=1307) (N=339)

SES
Low, % 50.5 39.2 10.3 3.8 71.9 24.3
Middle, % 52.7 36.0 11.3 7.8 71.2 21.0
High, % 59.5 33.8 6.7 8.1 73.8 18.1
Highest, % 58.8 34.4 6.8 10.9 72.8 16.3
Ethnicity
White, % 62.6 31.8 5.6 8.9 74.3 16.8
Black, % 35.5 46.6 17.9 5.8 66.8 27.4
Asian, % 54.3 40.0 5.7 7.8 77.1 15.1
Mixed, % 50.6 38.1 11.3 5.8 67.9 26.3
Tried smoking, %
No 90.0 79.5 76.7 92.3 82.6 73.4
Yes 10.0 20.5 23.3 7.7 17.4 26.6
Susceptibility to smoking, %
No 85.8 80.8 82.0 79.5 79.6 77.6
Yes 14.2 19.2 18.0 20.5 20.4 22.4
Weekend exercise, %
No days 33.5 36.5 40.0 30.7 24.8 16.3
1 day 43.1 43.0 37.3 26.3 33.2 35.3
2 days 23.4 20.5 22.7 43.1 42.0 48.3
Age, mean (S.D.) 11.80 (0.32) 11.90 (0.32) 11.98 (0.31) 11.72 (0.34) 11.80 (0.36) 11.89 (0.38)
BMI, mean (S.D.) 19.15 (3.56) 20.57 (3.73) 22.25 (3.67) 18.24 (2.75) 18.78 (3.21) 19.45 (3.58)
Days/week exercised, mean (S.D.) 3.70 (1.78) 3.63 (1.76) 3.94 (2.11) 3.97 (20.7) 4.19 (2.00) 4.33 (2.14)
High-fat food intake, mean (S.D.) 13.77 (4.90) 14.47 (5.10) 14.17 (5.34) 14.22 (5.41) 15.37 (5.47) 16.38 (5.77)
Fruit/vegetable intake, mean (S.D.) 4.24 (2.26) 4.29 (2.41) 4.39 (2.66) 4.00 (2.30) 4.03 (2.32) 4.41 (2.64)
Perceived stress, mean (S.D.) 5.70 (2.82) 5.79 (2.91) 6.47 (3.30) 5.34 (2.93) 5.29 (2.98) 5.54 (3.16)
Psychological diculties, mean (S.D.) 11.80 (5.04) 12.24 (5.07) 12.66 (4.85) 12.19 (4.87) 12.05 (5.14) 12.57 (5.36)

and girls. This was dicult to interpret for girls,
but again suggested in boys that susceptibility to
smoking increased with pubertal stage.
For exercise behaviour, a signicant relation-
ship was found between pubertal stage and
weekend exercise in boys. The more pubertally
advanced boys were more likely to take exercise
at the weekend. Although non-signicant the
results from the girls analyses suggested a re-
versal of this pattern with girls taking less
weekend exercise as they become more devel-
oped. Eects for overall days of exercise were
not signicant in either girls or boys, although
the results showed the same direction for boys.
For food intake, there was a signicant re-
lationship between pubertal stage and high-fat
food intake in boys with the more pubertally
advanced boys consuming larger amounts of
high-fat food, but no signicant relationship
between pubertal stage and high-fat food intake
in girls. There was no association between
pubertal stage and fruit and vegetable intake in
either boys or girls.
Mediation by stress and psychological diculties
For a variable to be a potential mediator it has
to be associated with both the independent and
the dependent variables (Baron & Kenny, 1986).
Mean scores for the stress and psychological
diculties scales across the pubertal groups are
shown in Table 2. There was a signicant as-
sociation between pubertal stage and stress
scores in girls, F(2, 1324)=3.50, P<0.05, and a
trend for psychological diculties to be higher
in those who were more pubertally advanced,
F(2, 1328)=2.08, P=0.13. The pattern of scores
for boys was neither linear nor statistically sig-
nicant, although the highest level of stress
and psychological diculties was in the most
pubertally advanced group.
1376 A. E. Simon and others

Table 3. Summary of regression analyses showing the associations between health behaviours and
pubertal stage in boys : Model 1 controlling for SES, age, ethnicity, BMI and Model 2 additionally
controlling for perceived stress and psychological diculties odds ratios (95 % CI )
Pubertal stage

N Pre Beginning Mid/Advanced

Model 1
Tried smoking 1244 1.0 2.39 (1.284.45)** 4.41 (2.188.91)***
Susceptibility to smoking# 1057 1.0 2.37 (1.125.04)* 2.55 (1.046.28)*
Days/week exercised 1259 1.0 1.22 (0.901.63) 1.31 (0.782.18)
Weekend exercise 1195 1.0 1.21 (0.891.65) 2.11 (1.343.33)**
Fruit and vegetable intake 1254 1.0 1.08 (0.791.48) 1.30 (0.961.75)
High-fat food intake 1216 1.0 1.54 (1.082.20)* 2.39 (1.523.70)***
Model 2
Tried smoking 1229 1.0 2.23 (1.174.26)* 4.17 (2.028.61)***
Weekend exercise 1184 1.0 1.24 (0.901.72) 2.19 (1.393.47)**
High-fat food intake 1203 1.0 1.68 (1.182.39)** 2.48 (1.633.86)***

# Among never smokers.

* P<0.05 ; ** P<0.01; *** P<0.001.

Table 4. Summary of regression analyses showing the associations between health behaviours and
pubertal stage in girls : Model 1 controlling for SES, age, ethnicity, BMI and Model 2 additionally
controlling for perceived stress and psychological diculties odds ratios (95 % CI)
Pubertal stage

N Pre Beginning Mid/Advanced

Model 1
Tried smoking 1603 1.0 2.36 (1.324.25)** 2.64 (1.096.41)*
Susceptibility to smoking# 1303 1.0 1.60 (1.082.34)* 1.37 (0.772.45)
Days/week exercised 1628 1.0 0.90 (0.731.13) 1.21 (0.712.03)
Weekend exercise 1576 1.0 0.79 (0.601.05) 0.60 (0.301.20)
Fruit and vegetable intake 1606 1.0 1.01 (0.711.43) 0.88 (0.521.49)
High-fat food intake 1574 1.0 1.22 (0.951.58) 1.23 (0.881.73)
Model 2
Tried smoking 1573 1.0 2.56 (1.454.54)** 2.42 (1.075.49)*
Weekend exercise 1552 1.0 0.80 (0.601.07) 0.62 (0.311.24)
High-fat food intake 1547 1.0 1.17 (0.921.51) 1.13 (0.771.67)

# Among never smokers.

* P<0.05 ; ** P<0.01.

To check for associations between stress or
psychological diculties and health behaviours
we used Pearson correlations, which showed
that they were signicantly associated with each
of the health behaviours that were associated
with pubertal stage, that is, having tried smok-
ing, weekend exercise and high-fat intake. Stress
was modestly but signicantly associated with
having tried smoking (r=0.09, P<0.001) doing
less weekend exercise (r=x0.06, P<0.05) and
higher intake of fatty foods (r=0.10, P<0.001)
in boys. The same patterns emerged for associ-
ations with psychological diculties (smoking,
r=0.18, P<0.001 ; exercise, r=x0.06, P<0.01;
intake of fatty foods, r=0.13, P<0.001). The
results were very similar for girls showing sig-
nicant associations with stress (smoking, r=
0.11, P<0.001 ; exercise, r=x0.08, P<0.01;
intake of fatty foods, r=0.10, P<0.001) and
with psychological diculties (smoking, r=0.16,
P<0.001 ; exercise, r=x0.07, P<0.01 ; intake
of fatty foods, r=0.14, P<0.001).
Given that the criteria for mediation were
met, logistic and ordinal regressions were car-
ried out controlling for stress and psychological
diculties to see if this reduced the variance
 Puberty and adolescent health behaviours
associated with pubertal stage (see Tables 3 and
4, Model 2). In the interests of symmetry, these
were done for both boys and girls, even though
the association between pubertal stage and
stress in boys was less clear cut. Looking at the
odds ratios and condence intervals for pubertal
stage, it is clear that including these factors,
although slightly changing the values, did not
have a substantial impact. Therefore, dierences
in stress and psychological diculties between
pubertal stages did not seem to be the key to
the variation in health behaviours linked with
puberty.
DISCUSSION
The results of this study support the hypothesis
that pubertal stage was related to health behav-
iour in a large sample of adolescents although
the patterning of results varied both by behav-
iour and gender. Being more pubertally ad-
vanced in boys was associated with a greater
likelihood of smoking and a less healthy diet,
but higher levels of exercise. In girls, more ad-
vanced puberty was linked with higher smoking,
but there was little association with either diet
or exercise. The association between pubertal
stage and health behaviours, where it was seen,
generally showed a graded eect, such that
the likelihood of engaging in a particular behav-
iour changed incrementally with increases in
pubertal stage.
The results on smoking status and suscepti-
bility to smoking replicated other ndings in this
area (Tschann et al. 1994; Gritz et al. 1998;
Harrell et al. 1998; Dick et al. 2000). For exer-
cise behaviour, advanced puberty was associ-
ated with more weekend exercise in boys but
not in girls. In girls there was the possibility that
weekend exercise may actually decrease with
pubertal stage, although this was a non-signi-
cant result. The results reect established gender
dierences in exercise participation among
adolescents (Lasheras et al. 2001). The results
also support the suggestion that higher pubertal
stage is associated with an increase in sedentary
activity among girls, but not boys (Bradley et al.
2000). The variable measuring days per week
exercised produced no signicant results. This
was probably because this variable is con-
founded with compulsory physical education at
school, which varies between schools and does
1377
not necessarily index voluntary participation in
exercise. Exercise outside school context may be
more important, as it indicates an adolescents
deliberate choice to exercise as part of their lei-
sure/social time. Pubertal stage was associated
with high-fat food intake only in boys in our
sample. The more developed boys may simply
need to eat more as they are older and taller.
Other studies have suggested that this gradient
in intake disappears after controlling for body
weight (Clavien et al. 1996) but in our data, the
eect persisted after controlling for BMI. These
ndings suggest that pubertal development in
boys is associated with an increasingly poor
diet. This increase in fat intake may be related to
the greater independence of older children to
make food choices as they decrease the amount
of meals eaten at home (Gillman et al. 2000).
There was no relationship between fruit and
vegetable consumption and pubertal stage in
either boys or girls, so although previous studies
have shown decreased consumption associated
with age (Vereecken & Maes, 2000), our data
suggest that this is not related to the pubertal
changes occurring at this time.
The main aim of the present study was to
test whether the associations between pubertal
stage and health behaviours were linked to pu-
bertys associations with either perceived stress
or emotional and behavioural diculties as
measured by the SDQ. A relationship was seen
between stress and pubertal stage in girls,
although no such relationship was found in
boys. Smaller eects between puberty and stress
in boys may have gone undetected due to the low
reliability of the PSS in this sample. However,
the SDQ also failed to detect a relationship,
although for both measures the most stress and
diculties were detected among the most ad-
vanced male puberty group. The boys were
overall less developed than the girls. The re-
lationship with stress and psychological di-
culties may only emerge among more advanced
puberty groups than the boys represented at this
age. Therefore, there may be value in exploring
this relationship further using longitudinal work
which tracks these groups over time. Having
noted these limitations, it was still clear from
the logistic regression analyses that these vari-
ables did not mediate the relationship between
pubertal stage and health behaviours in either
gender. On the basis of the results shown here,
1378 A. E. Simon and others
an alternative explanation for the link between
pubertal stage and health behaviours must be
found.
Although not tested here, it may be that other
social, rather than emotional, processes have a
role to play in explaining the health behaviour
dierences. Early maturation has been associ-
ated with social processes, e.g. socialization with
older peers, greater peer pressure, emulating
adult type behaviour (Silbereisen et al. 1989)
that may result in behavioural changes. Behav-
iours such as smoking and eating outside of
the home are associated with a more adult or
independent lifestyle and the changes in be-
haviour seen in this study could demonstrate
an attempt by the adolescent to assert their in-
dependence.
There were some limitations to the con-
clusions that can be drawn from this study. As
discussed, there were problems with the
measurement of perceived stress, which have
meant that the mediational hypothesis was not
tested as rigorously as it could have been. The
results from this study derive from cross-
sectional data which has meant that some
relationships have remained ambiguous. For
example, the results found here could reect
dierences in pubertal timing rather than dif-
ferences in levels of pubertal stage. That is, those
who are the most pubertally advanced at this
age, are also the group that have developed
relatively early compared to their peer group. As
the current pre/beginning/mid pubertal groups
progress through the higher pubertal stages they
may display similar behaviour to the current
advanced/post-pubertal groups, supporting a
maturational model. On the other hand those
who are currently more developed than their
peers might always stand out as a distinct group,
suggesting an eect of pubertal timing. It will be
possible to pursue this line of enquiry longi-
tudinally with future data from the HABITS
project.
This study was important for a number of
reasons. First, comparatively little research has
examined smoking uptake as part of a wider
issue of adoption of health behaviours such
as exercise and dietary choice. There is some
debate as to whether smoking is better con-
ceived as a health behaviour or is more closely
allied with problem/delinquent behaviour es-
tablished during this period (Turbin et al. 2000).
The analyses here have shown that dietary
behaviours in boys and exercise in girls are re-
lated to pubertal status in a similar way to
smoking behaviour, giving some support to the
health behaviour view. However, there was no
association with diet in girls and a positive as-
sociation with exercise in boys, suggesting com-
plex pathways best understood on a behaviour
by behaviour basis. Secondly, this study used a
large sample comprising both boys and girls.
Much research to date which considers pubertal
stage and behaviour has focused either on boys
or girls but not both. Incorporating both gen-
ders into the analyses has enabled a comparison
to be made between genders and there were a
number of important dierences. Thirdly, this
study provided the opportunity to explore
a possible causal explanation for the relation-
ship between puberty and health behaviour,
namely that puberty related increases in
emotional stress or distress were responsible
for the changes in behaviour. This has been ex-
plored in detail in only one previous study (i.e.
Tschann et al. 1994). The results here have
substantiated the conclusions of that study that
emotional distress does not mediate the relation-
ship between puberty and behaviour.
In conclusion, more advanced puberty was
found to be associated with changes in health
behaviours including smoking, exercise and diet.
The results suggest that maturity is a largely
unfavourable event for both boys and girls in
terms of the health behaviours that are estab-
lished during adolescence, with the notable
exception of exercise behaviour in boys. The
implications for those involved in health pro-
motion with adolescents, including school sta,
are that early-maturing adolescents need extra
support. This may be to help them withstand
exposure to new social pressures, or to come to
terms with the physical changes. The initiation
of these behaviours is probably unrelated to
issues of psychological stress and further work
could establish more clearly whether, and in
what way, pubescent adolescents are exposed to
unhelpful social inuences.
REFERENCES
Alsaker, F. (1996). Annotation : the impact of puberty. Journal of
Child Psychology and Psychiatry 37, 249258.
Baron, R. M. & Kenny, D. A. (1986). The moderator-mediator
variable distinction in social psychological research : conceptual,
 Puberty and adolescent health behaviours
strategic and statistical consideration. Journal of Personality and
Social Psychology 51, 11731182.
Benowitz, N. L. (1996). Cotinine as a biomarker of environmental
tobacco smoke exposure. Epidemiology Review 18, 188204.
Bradley, C. B., McMurray, R. G., Harrell, J. S. & Deng, S. (2000).
Changes in common activities of 3rd through 10th graders : the
CHIC study. Medicine and Science in Sports and Exercise 32,
20712078.
Brooks-Gunn, J., Warren, M. P., Rosso, J. & Garigulo, J. (1987).
Validity of self-report measures of girls pubertal status. Child
Development 58, 829841.
Caspi, A. & Mott, T. E. (1991). Individual dierences are accen-
tuated during periods of social change : the sample case of girls at
puberty. Journal of Personality and Social Psychology 1, 157168.
Clavien, H., Theintz, G., Rizzoli, R. & Bonjour, J.-P. (1996). Does
puberty alter dietary habits in adolescents living in a Western
society ? Journal of Adolescent Health 19, 6875.
Cohen, S. & Williamson, G. (1988). Perceived stress in a probability
sample of the United States. In The Social Psychology of Health
(ed. S. Spacapan and S. Oskamp), pp. 3167. Sage : California.
Cohen, S., Kamarck, T. & Mermelstein, R. (1983). A global measure
of perceived stress. Journal of Health and Social Behaviour 24,
385396.
Dick, D. M, Rose, R. J., Viken, R. J. & Kaprio, J. (2000). Pubertal
timing and substance use : Associations between and within
families across late adolescence. Developmental Psychology 36,
180189.
Gillman, M. W., Rifas-Shiman, S. L., Frazier, L., Rockett, H. R.,
Camargo, C. A., Field, A. E., Berkey, C. S. & Colditz, G. A.
(2000). Family dinner and diet quality among older children and
adolescents. Archives of Family Medicine 9, 235240.
Goddard, E. & Higgins, V. (1999). Smoking, Drinking and Drug use
Among Young Teenagers in 1998. UK National Smoking Surveys
Volume 1. The Stationery Oce : London.
Goodman, R., Meltzer, H. & Bailey, V. (1998). The strengths and
diculties questionnaire : a pilot study on the validity of the self-
report version. European Child and Adolescent Psychiatry 7,
125130.
Goodson, P., Evans, A. & Edmundson, E. (1997). Female ado-
lescents and onset of sexual intercourse: a theory-based review
of research from 1984 to 1994. Journal of Adolescent Health 21,
147156.
Gritz, E. R., Prokhorov, A. V., Suchanek Hudmon, K.,
Chamberlain, R., Taylor, W., Di Clemente, C., Johnston, D., Hu,
S., Jones, L., Mullin Jones, M., Roseblum, C., Ayars, C. &
Amos, C. (1998). Cigarette smoking in a multiethnic population
of youth : Methods and baseline ndings. Preventive Medicine 27,
365384.
Harrell, J. S., Bangdiwala, S. I., Deng, S., Webb, J. & Bradley, C.
(1998). Smoking initiation in youth : the role of gender, race,
socioeconomics and developmental status. Journal of Adolescent
Health 1998 23, 271279.
Hayward, C., Gotlib, I. H., Schraedley, P. K. & Litt, I. (1999).
Ethnic dierences in the association between pubertal status and
symptoms of depression in adolescent girls. Journal of Adolescent
Health 25, 143149.
Heath, G. W., Pate, R. R. & Pratt, M. (1993). Measuring physical
activity among adolescents. Centres for Disease Control and Pre-
vention Public Health Reports 108, 42S46S.
Laberge, L., Petit, D., Simard, C., Vitaro, F. & Tremblay, R. (2001).
Development of sleep patterns in early adolescence. Journal of
Sleep Research 10, 5967.
Lasheras, L., Aznar, S. & Merion, B. (2001). Factors associated with
physical activity among Spanish youth through the national health
survey. Preventive Medicine 32, 455464.
Lindquist, C. H., Reynolds, K. D. & Goran, M. I. (1999). Socio-
cultural determinants of physical activity among children. Pre-
ventive Medicine 29, 305312.
McCabe, M. P., Ricciardelli, L. A. & Baneld, S. (2001). Body
image, strategies to change muscles and weight, and puberty.
1379
Do they impact on positive and negative aect among adolescent
boys and girls ? Eating Behaviours 2, 129149.
Neumark-Sztainer, D., Story, M., Resnick, M. D. & Blum, R.
(1996). Correlates of inadequate fruit and vegetable consumption
among adolescents. Preventive Medicine 25, 497505.
Petersen, A. C. & Taylor, B. (1980). The biological approach to
adolescence : biological change and psychological adaptation. In
Handbook of Adolescent Psychology (ed. J. Adelson), pp. 117155.
Wiley : New York.
Petersen, A. C., Crockett, L., Richards, M. & Boxer, A. (1988).
A self-report measure of pubertal status : reliability, validity and
initial norms. Journal of Youth and Adolescence 17, 117132.
Pierce, J. P., Choi, W. S., Gilpin, E. A., Farkas, A. & Merritt, K.
(1996). Validation of susceptibility as a predictor of which ado-
lescents take up smoking in the United States. Health Psychology
15, 355361.
Ricciaridelli, L. A., McCabe, M. P. & Baneld, S. (1999). Body im-
age and body change methods in adolescent boys : role of parents,
friends, and the media. Journal of Psychosomatic Research 49,
189197.
Roe, L., Strong, C., Whiteside, C., Neil, A. & Mant, D. (1994).
Dietary intervention in primary care : validity of the DINE method
for diet assessment. Family Practice 11, 375381.
Rolland-Cachera, M.-F., Bellisle, F. & Deheeger M. (2000). Nu-
tritional status and food intake in adolescents living in Western
Europe. European Journal of Clinical Nutrition 54, 41S46S.
Silbereisen, R. K., Petersen, A. C., Albrecht, H. T. & Kracke, B.
(1989). Maturational timing and the development of problem
behaviour : longitudinal studies in adolescence Journal of Early
Adolescence 9, 247268.
Stattin, H. & Magnusson, D. (1990). Pubertal Maturation in Female
Development. Erlbaum : New Jersey.
Steinberg, L. & Sheeld-Morris, A. (2001). Adolescent development.
Annual Review of Psychology 52, 83110.
Sweeting, H., Anderson, A. & West, P. (1994). Socio-demographic
correlates of dietary habits in mid to late adolescence. European
Journal of Clinical Nutrition 48, 736748.
Tanner, J. M. (1962). Growth at Adolescence: With a General Con-
sideration of the Eects of Hereditary Environmental Factors upon
Growth and Maturation from Birth to Maturity, 2nd edn. Blackwell
Scientic : Oxford.
Tschann, J. M., Adler, N. E., Irwin, C. E., Millstein, S., Turner, R. &
Kegeles, S. (1994). Initiation of substance use in early adolescence :
the roles of pubertal timing and emotional distress. Health
Psychology 13, 326333.
Turbin, M. S., Jessor, R. & Costa, F. M. (2000). Adolescent cigarette
smoking : health-related behaviour or normative transgression ?
Prevention Science 1, 115124.
Van Wersch, A., Trew, K. & Turner, I. (1992). Post-Primary school
pupils interest in physical education : age and gender dierences.
British Journal of Educational Psychology 62, 5672.
Vereecken, C. & Maes, L. (2000). Eating habits, dental care
and dieting. In Health and Health Behaviour among Young
People, Health Behaviour in School-aged Children : A WHO Cross-
National Study (HBSC ) International Report (ed. C. Currie, K.
Hurrelmann and R. Smith), pp. 8385. World Health Organ-
ization : Copenhagen.
Wardle, J., Parmenter, K. & Waller, J. (2000). Nutrition knowledge
and food intake. Appetite 34, 269275.
Wardle, J., Robb, K. & Johnson, F. (2002). Assessing SES in
adolescents : the validity of a home auence scale. Journal of
Epidemiology and Community Health 56, 595599.
Wardle, J., Jarvis, M. J., Steggles, N., Sutton, S., Williamson, S.,
Farrimond, H., Cartwright, M. & Simon, A. E. (2003). Socio-
economic disparities in cancer-risk behaviours in adolescence :
baseline results from the Health and Behaviour in Teenagers Study
(HABITS). Preventive Medicine 36, 721730.
Williams, J. M. & Dunlop, L. C. (1999). Pubertal timing and
self-reported delinquency among male adolescents. Journal of
Adolescence 22, 157171.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.